seoFOMO's 2026 State of AI Search Optimization report landed this week. If you skim one thing in it, make it the section on what practitioners believe actually moves AI recommendations — because the answer is mostly off-site, and that changes what you can sell.
The short version
When an assistant answers "best [service] in [city]," it builds the response from third-party sources: roundup articles, directories, review platforms, forums, trade press, local news. The client's own site is the confirmation layer — it verifies who they are, what they do, and where. The recommendation itself is assembled from everyone else's pages.
Most agency AEO work so far has been on-site: schema, entity clean-up, FAQ blocks, content restructuring. Useful, necessary, and largely a one-off. The off-site half is the recurring half — and almost nobody is packaging it properly yet.
Packaging it
The unit of work is clean, which is what makes it sellable:
Phase one — the citation surface audit. Build a prompt set of 20–40 questions a real buyer would ask across the client's categories and service areas. Run them across the assistants that matter for that client. Log every cited source. You end up with an inventory of the 30–60 pages that actually decide who gets recommended in that market, plus a gap list showing where the client is absent, mentioned incorrectly, or mentioned alongside competitors but ranked below them. This prices comfortably as a paid discovery engagement.
Phase two — the placement retainer. A fixed number of placements or corrections per month against that gap list: listing claims and clean-ups, outreach to roundup authors, review velocity on the platforms that keep surfacing, supplier and association pages, local press. Re-run the prompt set monthly to show movement.
Bill it as its own line item. Do not absorb it into an existing SEO retainer. Folding it in gives the work away and quietly makes your SEO fee look inflated; naming it separately reopens the pricing conversation with a genuinely new deliverable attached.
Setting expectations before you sell it
The report is honest about measurement, and you should be too. AI referrals mostly land as direct traffic. Attribution is immature and will stay that way for a while.
So define the KPI up front as presence: how many of the client's priority prompts return them, how often they're cited, and how they place against named competitors. Then add one low-tech signal — a "how did you hear about us?" field on the client's intake — so there's a revenue-side number to point at by month three. Clients who agree to that framing before the first invoice do not churn over attribution later.
For your own shop
Two things. First, this work suits people you already employ — digital PR, outreach, local listings. The margin is good because the audit is systematised and the placements are junior-executable against a clear target list.
Second, run the prompt set on yourself. Prospects now ask assistants for agency recommendations, and the same rule applies: you get named because of directories, roundups, podcast appearances and client reviews, not because of your homepage copy.
SimpleAIO runs the prompt set, logs the cited sources and returns the gap list in plain English — so your team spends its hours on placements rather than assembling the audit. White-label reporting is available on agency plans.